SYSTEMS AND METHODS FOR INFERRING ENTITY RELATIONSHIPS VIA NETWORK COMMUNICATIONS OF USERS OR USER DEVICES
A number of techniques facilitate generation of data points from observations about network traffic. An inferencing system can use these data points to determine whether a relationship exists between two entities or whether an existing relationship has terminated, without any external knowledge of the existence of or termination of such a relationship.
This disclosure generally relates to techniques for cybersecurity assessment and, more specifically, to observing and analyzing network communications and to inferring relationships among entities that are associated with the users and/or user devices that are involved in the observed network communications.
BACKGROUNDAs organizations around the world, whether for profit, non-profit, or government rely increasingly on networked systems to obtain and to provide goods and services, security of an organization's computing and networking systems (“cybersecurity”) is becoming increasingly important. Techniques to audit an organization's systems, including both hardware and software systems, are generally known. In some cases, however, performing a security audit of the organization's systems may not provide a comprehensive or accurate assessment of the security of those systems. One reason is that an organization likely has business relationships with several other organizations, such as affiliates, parts vendors, technology service providers, suppliers, distributors, customers, clients, accountants, lawyers, government regulators, etc. Personnel from these other business entities may regularly communicate with and access the organization's systems and/or data using public networks and/or the systems of those other business entities, which may not be audited and/or may not be secure. As such, in some cases, the assessment of security of the systems of an organization of interest can benefit from the knowledge of which other organizations regularly access its systems.
Obtaining this information is not an easy or a straightforward task, however. The organizations themselves can be complex, having a hierarchical structure of several business entities, and one or more of those entities may form business relationships with other business entities. Moreover, the business relations can be dynamic. They may form at different points in time and may become dormant or may cease to exist after some time.
SUMMARYVarious embodiments described herein feature techniques and systems that can infer relationships among entities by analyzing network traffic associated with various entities. As the existence of a relationship between two entities is not known initially, the traffic is not monitored specifically, in a targeted manner, between a pair of entities. Rather, the traffic associated with several entities at large is observed. From the observed traffic, complementary occurrences of user devices or user identifiers are detected to be associated with assets, e.g., network or physical asserts, of two different entities. This detection can lead to an inference of the existence of relationship between those two entities.
In general, the system described herein implements various techniques for programmatically discovering possible relationships among organizations, and thus builds a business intelligence profile on arbitrary sets of organizations and companies. This information is valuable for risk management solutions as it identifies business partnerships without the need to engage with that company directly. This can be critical when compiling various risk metrics of organizations as these third-party relationships can have a significant impact on the operations of those organizations. These business relationships and business intelligence discovered using these methods may include:
-
- customers of an organization, including both B-2-C and B-2-B organizations.
- members of the organization's supply chain, for example such as those manufacturing parts, supplying parts, conducting designs, etc.
- business partners, for example such as those working together developing/marketing a shared product, completing a contract together, etc.
- information technology providers and cloud-based applications or hosting services;
- one company is in the process of evaluating or conducting an acquisition of another company; and
- any other reason why two organizations may be associating with one another.
Some techniques may not be able to identify the specific type of business relationship between the two organizations, but can infer the possibility that a business relationship exists. The fact that the possible business relationship has been identified, and the parties identified, may then allow the type of business relationship to be inferred. For example, if a business relationship has been identified between a defense contractor and a company that specializes in developing firmware for GPS receivers, a manufacturer-supplier relationship may be inferred. Likewise, if a relationship is inferred between a defense contractor and a hydrostatic pump testing facility, a consumer-service provider relationship may be inferred. The strength of the business relationship may be inferred by factors such as the duration of the relationship appearing through the techniques described herein, as well as whether the relationship appears through more than one techniques.
In various embodiments, an inferencing system described herein can use the following techniques, either individually or in combination for discovering relationships. These include the following.
Technique 1—Observing and cataloging the activity and patterns of devices accessing the Internet through organizational networks, and thus associated with an IP address revealed through cookies, device fingerprints, or unique identifiers associated to the device. By mapping the IP addresses that these cookies, fingerprints, or device identifiers have appeared on to the organizations or companies responsible for those network resources, a determination can be made as to which organizations that user or device is associated with. Modeling the characteristics of those visits to these organizations facilitates the inference of relationships between those organizations. Examples of data sources include but are not limited to datasets derived from web site operators, ad technology, marketing technology, email and other forms of common and uncommon forms of transactions.
Technique 2—Observing and cataloging the activity and patterns of devices accessing the Internet through organizational networks to a particular service, and thus having an association with an IP address which are then revealed through user account information or other user identifiers or user-like identifiers specifically provided by the user or device. These may include, for example, email readability metrics (such as a system as described in U.S. application Ser. No. 16/015686, attached as Appendix A), email address, or phone number. For example, a user logging into a service of any protocol or application (e.g. IMAP, POP3, XMPP, or web-based services) from multiple IP addresses implies that the user had access to that network with the corresponding IP address. Similar to modeling on characteristics of the device, one may look at the characteristics of the user behavior to understand that relationships exist and of which type.
Technique 3—Dissecting network flow data and identifying the two parties for a given network transaction. By mapping the two IP addresses that are communicating with each other and referencing other characteristics about that transaction (protocol, port, length of session, etc.), one can infer a relationship of various certainty. Various protocols establish a stronger relationship between companies than others, such as SMTP (email) and XMPP (instant messaging), while others might reveal a relationship but is likely to be weaker, such as HTTP/S (web site visits) and FTP (file transfer).
Technique 4—Information made publicly available by various malware detection software applications and hardware appliances can reveal the parties of a particular email or message. This may include various combinations of: the source party via its originating MTA or device IP address, source domain, and/or source email address combined with the destination party via one of its receiving MTA IP addresses, destination domain, and/or destination email address. By cataloging such information and mapping the email domains to organizations and companies, an index and database of companies can be compiled that reveals communications among the companies. Additional filtering can be performed to selectively choose individuals at a company that might reveal a stronger relationship compared to another, such as a CEO, if the contextual information such as revealed email addresses can be joined with an external dataset.
Technique 5—Precise GPS coordinate data sets that contain IP address information or a device or user identifier can reveal users employed by a specific company at the physical location of another company or organization, and thus a possible relationship might exist between the two. This is a common occurrence for situations where a user has a VPN enabled on their device and thus their publicly visible IP address is their company's network, but the user's physical location is at a different location. By using geo-based APIs, one can reverse the geo-coordinates to a physical address and subsequently map this to an organization.
One or more of these techniques may be used to discover new potential relationships, to maintain known relationships, and to terminate known relationships. Embodiments of an inferencing system can use a variety of techniques to increase confidence in relationships, defined as company pairs, as well as to infer the type of relationship that exists.
Such an inferencing system can work in concert with systems that map IP addresses to businesses and organizations, such as those defined in U.S. Pat. No. 9,438,615, incorporated herein by reference in its entirety. Other mechanisms may also be used to determine the authoritative ownership of various forms of digital assets (e.g., domains, IP addresses, etc.) such as using direct sources (e.g. DNS, WHOIS data from RIRs, etc.), or those derived from such sources (e.g. firms or services extracting this information).
Embodiments of an inferencing system using these techniques can apply the aforementioned techniques to model relationship strength and currency, as well as to apply ultimately a confidence value based on the respective companies' industries, products and services, and/or frequency of appearance in the data set(s).
Accordingly, in one aspect a method is provided for inferring a relationship between two entities. The method includes the step of receiving at a server a network observation dataset, where each data point in the dataset includes (a) a device identifier and, additionally, (b)(i) a network address or (b)(ii) a device location. The method also includes: selecting from the dataset a first-type data point that includes: (A) a first device identifier, and (B)(i) a first network address or (B)(ii) a first device location; and identifying a first entity associated with the first network address or the first device location. The method further includes: selecting from the dataset a second-type data point that includes: (A) the same first device identifier, and (C)(i) a second network address different from the first network address or (C)(ii) a second device location different from the first device location; and identifying a second entity that is different from the first entity and that is associated with the second network address or the second device location. The method further includes inferring a relationship between the first entity and the second entity. The inference of the relationship is associated with a baseline confidence level.
The first network address may include: a first Internet Protocol (IP) address, a first service set identifier (SSID), or a first basic SSID (BSSID), and the second network address may include a second IP address, a second SSID, or a second BSSID.
In some embodiments, the method includes selecting from the dataset a third-type data point that includes: (D) a second device identifier, and (C)(i) the second network address or (C)(ii) the second device location; and selecting from the dataset a fourth-type data point that includes: (D) the same second device identifier, and (B)(i) the first network address or (B)(ii) the first device location. The method may include increasing the baseline confidence level when the third-type and/or the fourth-type data points are selected.
In some embodiments, the method may include determining that: (i) a total number or frequency of the first-type data points in the dataset is at least equal to a specified home-count or home-frequency threshold; and also determining that: (i) a total number or frequency of the second-type data points in the dataset is at least equal to a specified visitor-count or visitor-frequency threshold. The method may include increasing the baseline confidence level when the above-described determinations are made.
In some embodiments, the method includes receiving an enrichment attribute and testing whether the enrichment attribute is associated with the first entity or the second entity. In these embodiments, the method also includes increasing the baseline confidence level if the enrichment attribute is determined to be associated with the first entity or the second entity, and otherwise, decreasing the baseline confidence level or invalidating the inferred relationship. The enrichment attributed may be included in the first data point or the second data point.
The enrichment attribute may include an entity identifier, and the testing step includes determining whether the entity identifier identifies the first entity, the second entity, or a different third entity. Additionally or in the alternative, the enrichment attribute may include a service set identifier (SSID), and the testing step may include determining whether the SSID is associated with the first network address or the second network address or neither the first nor the second network address. Additionally or in the alternative, the enrichment attribute may include a physical location associated with the first or the second network address.
In some cases, the enrichment attribute may include a classification of the first or the second network address, where the classification may indicate whether the corresponding network address is a residential address, an address associated with a cellular network, a business-related network address, etc. The testing step may include determining whether the classification indicates a business entity.
In some cases, the enrichment attribute may include a list of Access Points (APs) associated with the first network address, and the testing step may include determining whether one or more APs in the list are designated as non-guest or employee APs. Alternatively or in addition, the enrichment attribute may include a list of Access Points (APs) associated with the second network address, and the testing step may include determining whether one or more APs in the list are designated as guest or visitor APs.
In some cases, the enrichment attribute includes a list of devices associated with the first or the second network address and the list also includes a respective type of each device indicating, e.g., where the device is an endpoint device or sever device. The testing step may include determining whether one or more devices in the list are endpoint devices. If no devices in the list are endpoint devices, it is likely that the first or the second data points are erroneous.
In another aspect, a system is provided for inferring a relationship between two entities. The system includes a processor, and a network port in communication with the processor and adapted to receive from a data source a network observation dataset. Each data point in the dataset includes one or more of: (i) a device identifier, (ii) a user identifier, (iii) a network address, or (iii) a device location. The system also includes a memory coupled to the processor and having stored therein instructions which, when executed by the processor, program the processor to: select from the dataset a first-type data point that includes: (A)(i) a first device identifier or (A)(ii) a first user identifier and, additionally: (B)(i) a first network address or (B)(ii) a first device location. The instructions also program the processor to identify a first entity associated with the first network address or the first device location;
In addition, the instructions program the processor to select from the dataset a second-type data point that includes: (A)(i) the same first device identifier or (A)(ii) the same first user identifier and, in addition: (C)(i) a second network address or (C)(ii) a second device location. The instructions further program the processor to identify a second entity associated with the second network address or the second device location, and infer a relationship between the first entity and the second entity. Furthermore, the instructions program the processor to designate to the inferred relationship a baseline confidence level.
In some embodiments, the network port is adapted to receive an enrichment attribute, and the instructions further program the processor to: test whether the enrichment attribute is associated with the first entity or the second entity, and to increase the baseline confidence level, if the enrichment attribute is determined to be associated with the first entity or the second entity and, otherwise, decrease the baseline confidence level or invalidate the inferred relationship.
The enrichment attribute may include: (i) an entity identifier, (ii) a service set identifier (SSID), (iii) a physical location associated with the first or the second network address, (iv) a classification of the first or the second network address, (v) a list of Access Points (APs) associated with the first or the second network address and a respective type of each AP, or (vi) a list of devices associated with the first or the second network address and a respective type of each device. The enrichment attribute may be included in the first or the second data point.
In another aspect, a method is provided for inferring a relationship between two entities. The method includes the step of receiving at a server, a network observation dataset, where each data point in the dataset includes a user identifier and a network address. The method also includes selecting from the dataset a first-type data point including a first user identifier and a first network address; and identifying a first entity associated with the first network address. The method further includes selecting from the dataset a second-type data point that includes the same first user identifier and a second network address different from the first network address; and identifying a second entity that is different from the first entity and that is associated with the second network address. In addition, the method includes inferring a relationship between the first entity and the second entity, and associating a baseline confidence level with the inferred relationship.
In some embodiments, the method further includes selecting from the dataset a third-type data point including a second user identifier and the second network; and selecting from the dataset a fourth-type data point including the second user identifier and the first network address. The method may also include increasing the baseline confidence level when the third-type and/or the fourth-type data points are selected.
The method may include determining that: (i) a total number or frequency of the first-type data points in the dataset is at least equal to a specified home-count or home-frequency threshold; and determining that: (i) a total number or frequency of the second-type data points in the dataset is at least equal to a specified visitor-count or visitor-frequency threshold. The method may further include increasing the baseline confidence level when both above-described determinations are made.
In some embodiments, the method includes receiving an enrichment attribute; testing whether the enrichment attribute is associated with the first entity or the second entity; and increasing the baseline confidence level, if the enrichment attribute is determined to be associated with the first entity or the second entity and, otherwise, decreasing the baseline confidence level or invalidating the inferred relationship. The enrichment attribute may include: (i) an entity identifier, (ii) a service set identifier (SSID), (iii) a physical location associated with the first or the second network address, (iv) a classification of the first or the second network address, (v) a list of Access Points (APs) associated with the first or the second network address and a respective type of each AP, or (vi) a list of devices associated with the first or the second network address and a respective type of each device. The enrichment attribute may be included in the first or the second data point.
The present invention will become more apparent in view of the attached drawings and accompanying detailed description. The embodiments depicted therein are provided by way of example, not by way of limitation, wherein like reference numerals/labels generally refer to the same or similar elements. In different drawings, the same or similar elements may be referenced using different reference numerals/labels, however. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating aspects of the invention. In the drawings:
Objectives of the systems and techniques described herein include the discovery and description of business-entity relationships amongst various commercial and other organizations. The data gathered by monitoring and analyzing network activity and/or communications, as described herein, may be used for many purposes, such as to understand the third-party ecosystem of a particular organization. This can be important to a security rating service tasked with assessing and/or ranking the strength or effectiveness of an organization's cybersecurity practices, as the overall cybersecurity risks at an organization can be affected by the cybersecurity practices at various third parties with whom the organization communicates regularly or frequently. Understanding the organization's technology ecosystem can also be important to other organizations operating with, considering, or evaluating a potential relationship with the organization. For example, the organization's cybersecurity data and rating allows for a more accurate view into potential effects of a breach, downtime event, or other business-impacting event, it allows for the continual monitoring of companies for third-party risk management purposes, they can be used to construct accurate risk assessment models for a given service, industry, or function type to support issuance of cyber-insurance, evaluation of security posture, or a probability of breach. This information can also be useful and important for understanding the relationships among organizations including general insurance firms, financial firms, government regulators, IT service providers, infrastructure providers, organizations performing competitive research, market intelligence firms, among others.
The techniques described herein each represent approaches to inferring business relationships initiated by the unintended behaviors typically performed by users of devices, or of the devices themselves. Some of these devices may not need an attentive user interacting with the device to instigate the data point that allows for a relationship to be constructed within the system. Some data points may be generated via programmatic APIs, services, SaaS platforms, application service providers, on-premise software, or other systems, where an explicit user action may not have occurred or may not be necessary to reveal the relationship.
Each technique describes the creation and collection of various data points that can be consumed by the same common inferencing system for processing data points that ingest data about business relationships. The inferencing system is agnostic to the source, although the system may consider the source and the technique for making decisions in terms of confidence that the data point is meaningful, for determining reliability of the data points including historical and future observations, and may consider other factors that influence the confidence of a particular business relationship built upon data points generated using these techniques. The inferencing system is not limited to these techniques, and in addition to using solely those described herein, can employ other techniques previously documented or implemented, such as those outlined in U.S. Pat. No. 9,438,615, the entire contents of which are incorporated herein by reference.
Each technique may reveal different types of business relationships, as the inferencing system and processing data points may reveal relationships that are generally agnostic to the specific form of relationship between two organizations or how that information is revealed, although, such additional information can be used as an input or indicator when determining the confidence of a particular relationship.
The techniques described herein optionally involve additional steps during or after ingesting data points generated by these techniques, that focus on increasing reliability of a given business relationship derived using these methods, each of which is described below. However, these techniques can be enhanced by understanding relationships derived through other known techniques, for example, to increase confidence in other plausible relationships, or for relationships that are already known to be accurate by an inferencing system. Such relationships can be used, for example, in a supervised machine learning model as inputs based on the data sets, thus training the system to favor transactions and activities that lead to higher confidence results.
Technique 1: Inferencing Based on Device ID and IP Addresses AssociationThis inferencing technique for identifying a possible business relationship between two entities is based on assumptions and observations that devices, or the owners of those devices, may have an implicit or explicit relationship with or dependency on the organizations that provide and operate the networks to which those users and/or their devices are connected. In general, data points that contain identifiers associated with devices are collected as information indicating the IP address the devices use, as viewed by another device on the Internet. These data points are analyzed to understand the networks and the connection behavior of those devices on different networks through time. The data set typically includes a timestamp, an identifier for the device, which may be associated to an application on the device, as well as the IP address of the device as viewed externally. Other information about the device, its network, or other contextual information may be included within the dataset. Examples of such information include the device's internal (LAN) IP address, Dynamic Host Configuration Protocol (DHCP) lease time, etc. This additional information may also include multiple device identifiers and/or the information collected according various other entity relationship inferencing techniques that are described herein.
The identifier of a device may take different forms, either individually or in combination with other identifiers, including but not limited to a unique device identifier as provided by the operating system of the device (such as a UDID), an identifier provided by the manufacturer (such as a serial number), an identifier provided to a browser on the device (e.g., using a cookie), an identifier as provided by a mobile application on the device to the application itself (an application-specific unique identifier), an identifier associated to a device modem (such as an IMEI number), an identifier associated to the network interface (such as a MAC address), an identifier associated with the SIM card (such as the SIM number), an identifier associated with the telephony services (such as a phone number), an international mobile subscriber identity (IMSI) number, an identifier associated with a hotspot (such as a SSID and or BSSID). A form of identifier associated with a device as provided in the data set may be generic and may be present in other unrelated devices that are not in the data set. In some cases, the form of identifier may be unique to the data set itself, e.g., a salted hash of one or more other identifiers. In general, the choice of the form of the device identifier has no significant bearing on the quality or integrity of system is producing accurate results.
Some identifiers may span multiple devices, such as a telephone number that is associated to both a mobile phone as well as a laptop, or multiple mobile devices, particularly if the phone number is associated with a VoIP service. Some identifiers, such as cookies or phone numbers, may be logically applied to the device or to the user by the application that provided the cookie, but this technique is generally agnostic to the use of the application. The identifier associated with the device may be perpetual or it may be time-limited, but it may only be valuable if it is distinct from all other devices, and if it is visible on two or more data points. In general, at least two data points are needed because the same device ID must be seen on two networks that are respectively associated with two different entities—so that a relationship can be inferred between those two entities.
The data set may contain the IP address assigned locally to a device which, in some embodiments, may be an RFC1918 IP addresses In general, however, some networks may assign a publicly-routable IP address directly to the device, where the IP address can be associated with an organization. This can be beneficial and accurate, particularly if other observer devices on the Internet view the device of interest transgressing through an explicit proxy where the IP address assigned to the device is changed to that of the proxy. For example, when a device has an RFC1918 assigned address, generally it can only communicate with other devices on the Internet if it uses a Proxy or Network Address Translation (NAT) function. This will change the IP address associated with the data communication to be a public IP address. In many cases, this IP address also belongs to the same company, i.e., the company of interest.
The IP address may be the form of an IPv4 address or an IPv6 address. An IPv6 address may allow for noise reduction if the device is connected to multiple networks that use SLAAC and embed the device's MAC address into the IPv6 address. The IP addresses for the different data points corresponding to a particular device can be independent of each other. Also, whether the device has previously connected to an IPv4 network and subsequently connects to an IPv6 network or vice versa has no bearing on the integrity or quality of the data. Only when a device connects to two or more networks supporting SLAAC for IPv6 does the IP address itself have a bearing into reducing noise in the network.
The user device may be any Internet-capable device, such as a laptop, desktop, mobile phone, Internet of Things (IoT) device, etc. Neither the age of the device, nor the age of the operating system running on the device, has any significant bearing on the quality of the data received from or the observations made about the device according to the above-discussed attributes. The device may or may not be used often by one individual, and may be used by many individuals. Likewise, the device may not be used by any person, other than the possibility of the device being serviced by a person. The actions of the software on the device may be responsible for the generation of data points within the data set.
The originator of data points that form the dataset may take many forms, and data points in many different forms may be combined prior to analysis thereof by an inferencing system without the need to differentiate between those forms. The different forms include the information collected by the developers of various mobile applications that are installed on the device, information collected by the operators of web applications and websites, information collected by third-parties such as that collected by various add-ons or libraries embedded within the applications or websites, information collected by mobile or telephone operators, and many others that may understand or know at a certain time what an identifier associated with a device is as well as what the IP address associated with the device is. When both of these are known by any observer, including the time of observation by that observer, the observed information may be included in a data set used for relationship inferencing described herein. The observed data points may be received at a computer/server analyzing these data points periodically or in an on-demand manner, or when a particular system making the observations transmits those observations/data points to the computer/server analyzing the data points.
With reference to
At step “B1”, the device (Laptop 1) connects to a remote system (e.g., a web service or a web application) via the Internet. This action may have been initiated by the user, by the device itself, by a software operating on the device, and/or may be initiated in another form that would cause the device to seek an Internet transaction. During this transaction the device provides the remote system with an identifier that uniquely identifies the device. Such an identifier may have been provided to the device by the local system providing the network to which the device is connected. In some cases, the device may connect to the remote system more than once, and the remote system may allocate a unique identifier to the device. The device identifier may also be provided by a third party, such as the device manufacturer, or the maker of the operating system installed on the device. In general, the device identifier can be any of the several types of identifiers that are described above. Also, the device need not be a laptop. It can be any device, such as those described above, that can connect to a network.
An observation made by the remote system such as a particular device having a particular device identifier connecting at a certain time to a network and was allocated a particular IP address controlled by a certain entity, represents one data point. In some cases, a transaction between a client, i.e., the device of interest, and the remote system comprises the data point(s), rather than the transactions themselves. For example, a mobile application may periodically collect information about device identifiers and assigned IP addresses, and report it to a remote system.
At step “C1,” the device (Laptop 1) has connected to a different network, in this case ultimately belonging to “Business Partner, LLC”, which is responsible, controls, uses, and allocates to the devices on its network different IP address(es) than those used by “Example, Inc.” However, the previously provided identifier remains the same in its earlier and follow-up transaction with the remote system.
At step “D1”, the device (Laptop 1) returns to the previous network and initiates another transaction with the remote system. Likewise, at a later point the device performs another transaction with the remote system at step “E1”. From these observations, the inferencing system may determine that a particular device connects to the network(s) of Entity “Example, Inc.” frequently, and that the same device also connects to the network(s) of “Business Partner, LLC.” Based on these determinations, the inferencing system may infer a business relationship between the two entities “Example, Inc.” and “Business Partner, LLC.”
The transactions discussed above may result in example data points shown in
In some cases, the identifier for the device may change at some time between different transactions, as shown in the example data set depicted in
With reference to
In “C2” this transaction shown at “B2” is repeated. At “D2”, the other device (Laptop 2) is connected to the network controlled, used, and/or operated by the “Example, Inc.” This is determined from the IP address allocated to the other device. During this stage, the other device makes a transaction with the remote system while providing it with the same identifier, but the new IP address. At “E2” the other device (Laptop 2) returns to the original network.
These transactions may result in the data points that are shown in
Note that in D2, the IP address known to be associated with “Example, Inc.” is different from the IP address known to be associated with the same entity “Example, Inc.” in
The observable patterns discussed with reference to
In some cases, there may exist additional metadata about the network that can provide more contextual information into the networks to which the devices connect. Likewise, there are a number of data attributes that can improve confidence in the individual data points or when drawings an inference. Various such metadata and data attribute that can enrich the observed data points and inferences drawn therefrom, in general, are described below. There are specific types of enrichments that can benefit data points generated according to the processes described with reference to
In particular, having knowledge of the wireless networks associated with each IP address can provide information about whether the user might be on a guest network of an organization, increasing the value of that particular data point. For example,
Users transitioning across multiple guest networks at different times, e.g. one data point at each, may not necessarily be sufficient to establish a relationship considering the user or device that created these data points might have been visiting different companies, and was not associated with or operating at any of those entities. As such, several data points across a period of time are generally necessary to understand reliably the “home” organization associated with a user or a user device. For example, with reference to
Determining the networks to which the devices typically connect can play an important role to solidify further the confidence, because the chance that a relationship exists increases if there are bi-lateral observations of devices connecting to two opposing networks or groups of networks. For example, if a device that is typically observed to be active on Network A visits Network Z for a brief period of time and returns to Network A, while another device that is typically observed to be active on Network Z visits Network A for a brief period of time and returns to Network Z, these observations demonstrate a bi-lateral relationship in the context of this data set. It is not necessary to have this observation of data points to conclude that a relationship exists, but such observations tend to increases the confidence that such a relationship may exist, as discussed above with reference to
Another benefit that is gained by understanding the typically observed networks of a device is that one can disambiguate between activity profiles of devices, for example, to identify roaming device that visits many different networks, and is not related to any of those networks besides its home networks. For example, if a device traditionally is seen on Network A, and then is observed visiting Network Y briefly, and also briefly visiting Network Z, there is a lower probability of a relationship between Network Y and Network Z than there is between Network A and Network Y, or between Network A and Network Z. The more time a device has spent on a particular network is generally the network for which relationships with higher confidence may be inferred.
There is also no dependency for a specific device to revisit the same network of an organization in the future in order to conclude the re-observation of a relationship between organizations. It can be any other device that would also independently conclude a relationship between the organizations in question. The number of distinct and different devices that are observed in the data set that exhibit a relationship can also increase the confidence in the inference of a relationship.
The technique described above does not have a strong dependency on identifier lifetimes because it is agnostic to the particular device or the user. The device identifier lifetimes may be used, however, to derive an accurate population of devices. In general, the larger the population (i.e., the number of devices indicating the existence of a relationship) as a function of the entity size the higher the confidence of the interdependency or the inferred relationship between the two entities.
There are a number of factors and attributes about data points generated through the technique described above that can lead to noise, increase the likelihood of a data point being a false positive, or lead to a false positive conclusion about the existence of a relationship. One such example includes data points that have short-duration device identifiers associated to devices, such as those that rotate frequently. With such identifiers, the confidence in the inference drawn may be low because discovering other corresponding data points may be less likely. It would also be difficult to understand the devices' profile. As such, an inference system processing such a data set may need to observe data points over a longer time (e.g., a certain number of days or weeks) before considering data points various devices for relationship inference. Thus, in some embodiments, data corresponding to the devices having short-duration identifiers may be excluded or weighted less in favor of data corresponding to devices having a long duration or permanent identifiers.
Technique 2: Inferencing Based on User ID and IP Addresses AssociationThe second technique that an inferencing system identifying possible business relationships may use is based on the assumption and observation that users, or individuals themselves, who may operate one or may devices, may have an implicit or explicit relationship or dependency with the organizations that operate the networks to which those users or their devices connect. This technique uses data points containing identifiers associated with the user, as opposed to user devices, combined with information indicating which IP address the user's device uses as viewed by an external device on the Internet, and is based on subsequently understanding the networks and behavior of those users on different networks through time. In addition, the user identifier may contain extra information that implies the associative organization or other contextual information about the primary context of the user.
Each data point in the data set required for this technique generally includes a timestamp; a form of identifier associated with the user, which may be associated to a digital account that user operates (e.g. email address), or other identifier commonly limited to one person (e.g. phone number), or something intrinsic to the user (e.g. the user's name and zip code); as well as the IP address of the user device as viewed externally. Other information about the device, its network, or other contextual information based on the technique for which data points are extracted may be contained within the dataset, and such additional information can be useful to a system inferring business relationships. This extraneous information may be device identifiers in addition to the user identifier, additional information about the user, or it may be information used in connection with the other inferencing techniques described herein.
Like Technique 1, this technique also relies upon data points having IP address information. In general, an IP address in a data point may not be classified as an RFC1918 address, except where it is provided alongside the device's closest routable IP address to help with disambiguation when more than one user access a service under the same user identifier, or with other means to better manage volume. The IP address may be in the form of an IPv4 address or an IPv6 address, and the IP addresses for the different data points associated with a particular device may be independent of each other. Thus, whether a particular device previously connected to an IPv4 network and subsequently connects to an IPv6 network or vice versa has no bearing on the integrity or quality of the data. Networks configured to use SLAAC when using IPv6 addresses can provide additional information that may be used as a device identifier.
This technique is generally agnostic to the source of the data, where one type of service accessed by the users is not necessarily preferable to another. It is plausible that the data from various services is additive in its benefit since user identifiers are expected to remain constant across different services and service providers. For example, a user identifier that is an email address, “user1@example.com” will be the same across different services, but a device identifier that is a cookie is expected to be different across services, unless those services are controlled by a single party, and use the same cookie across such services. A user may have more than one type of identifier applied to that user, such that “user1@example.com” and “alice@example.com” are the same individual, but this technique is agnostic to that possibility and may treat these user identifiers as being associated with different users. An additional database can be used to determine, however, that “user1” and “alice” refer to the same person.
This technique is also generally agnostic to the action that triggered the transaction between the user and the service. Each service may provide the user different ways in which the users can interact with the service, such as through a web application over HTTPS, a mobile application, or third-party client over any other applicable protocol, such as IMAP, POP3, SMTP in the case of email service providers, etc. In one or more circumstances, the service may have the ability to identify the user and, if the service is capable of performing user identification, the particular user interaction with the service may produce a viable data point. This technique is also generally agnostic to the type of the device the user used for interacting with the service. For example, whether one or more data points are generated by users using mobile devices, or desktops, or laptops has no substantial bearing on the inferences drawn using this technique.
In addition, the user may perform different actions with the service, such as logging in or performing another action (e.g., querying information from a database; issuing a control command, etc.), the service supports. This technique is generally agnostic to what the individual is doing with the service other than that the service must be able to identify the user for the observed data points to be valuable. Lastly, the user themselves may trigger an action, or a programmatic or automatic process on the device may have caused a transaction to occur with the service. If the programmatic process or automatically performed a transaction (e.g., synchronizing local data on a user device with server data; reporting device status to a server, etc.), on behalf of the user and, as such, identified to the service the user, this transaction may be as viable as a user-initiated action for generating a data point. As such, any interaction with the service for which the user is identified from the service's perspective may provide as a valuable data point.
This technique can also analyze data points generated by many different types of services, and the data produced by any service that can at least identify a user as described herein. Examples of services include a variety of websites and web applications providing different functions, email providers, chat applications, among others. This also includes specific actions within other services such as the action of users reading individual emails (i.e., email readability metrics) as described in US 2018/0375822A1, the entire contents of which are incorporated herein by reference. US 2018/0375822A1 describes a system used for discovering IP addresses and domains that are controlled and operated by various organizations, and this data may also be used to infer business relationships as described herein.
At “B1,” the user causes the user device to interact with a remote system that is accessible via the Internet and that provides a particular service. For example, the user may log in into the service. This action may have been initiated by the user, by the device itself, any software operating on the device, and/or any other operation that would cause the device to transact with the remote system. Within this transaction, the device provides the remote system with a user identifier, e.g., an email address including the username for the service. Note that the action in “B1” or any subsequent action can be any action that is functionally different from logging in, as long as the service can identify the user via a supported identifier.
Thereafter, at “C1,” the user connects the user device to the network provided, controlled, and/or operated, by “Business Partner, LLC.” While so connected, the user interacts with the service provided by the remote system by logging in. Here again, this transaction may include the same user identifier (e.g., the email address) but a different IP address would be associated with the user device. A new data point may thus be generated, associating the user identifier with an IP address associated with another entity (i.e., “Business Partner, LLC”). By comparing the data points observed at “B1” and “C1” it may be inferred that a relationship exists between “Example, Inc.” and “Business Partner, LLC.” Additional transactions, as well as transactions that extend over a long period of time (e.g., hours, days, weeks, etc.), and/or one or more other factors described below, can increase confidence in the inference of the relationship.
For example, at “D1,” the user transitions back to the network provided, controlled, and/or used by “Example, Inc.” and may transact again with the remote system providing service. The user may repeat this again at “E1.” The sequence of transactions described above would result in the dataset that is shown in
In order to derive relationship information from this dataset, the IP addresses allocated to the user device(s) must be mapped to organizations. In cases where the user identifier provides additional metadata, such as having a domain contained within the user identifier, that domain may be mapped to an organization. If the associative organization is confirmed not to be a general provider of services or an email service provider, then it can be used as an input to understand the primary-associated organization of the user, which could plausibly be corroborated by the information gathered through IP addresses themselves. Making associations between domain names and IP addresses is described in US 2018/0375822A1, and those techniques may be used in various embodiments.
Notionally, if the user identifier is in the form of an email address, the inferencing system can aggregate the data points observed for all users for a given domain and treat the corresponding user identifiers as one common identifier for the purposes of processing the data. However, since each user may have his or her own unique characteristics and may use the services in different ways, it is important to have an understanding of all users who are associated with which organizations. As such, performing the data reduction described above may be avoided in some embodiments. The entropy provided by each user and service may also help with machine learning processes built for analyzing this dataset.
Technique 3: Inferencing Based on IP Addresses and Protocol AssociationThe third technique generally involves inferring entity relationships is based on the assumption and observation that two entities may have a business relationship if the users or user devices associated with those entities are observed to be interacting with one another via certain communication protocols, the use of which may be determined from the port numbers used in the communication. As such, this technique uses data points that generally contain information about two devices, or more specifically two IP addresses, one belonging to one organization and another IP address belonging to another organization, as well as contextual information about the ports, and thus implied services and/or protocols. Other metadata about the information exchanged may also be included in one or more data points.
Thus, each data point in the data set used by this technique generally includes a timestamp, two IP addresses of the two systems or networks communicating with each other, and TCP or UDP port of the system with which the connection was originally established. One or more data points may include additional information such as the TCP or UDP port of the system that initiated the connection, the number of packets sent, the total size of the data exchanged, etc. Such information is commonly referred to as “flow” information that describes packet flows within networks.
The data points used in this technique generally do not use RFC1918 IP addresses because such an IP address usually does not provide any notional value as a standalone attribute. An RFC1918 IP address may be provided alongside a publicly routed IP addresses where the RFC1918 IP address can provide value disambiguating between multiple clients behind a NAT. The IP address may be the form of an IPv4 address or an IPv6 address. An IPv6 address may allow for noise reduction if the device is connected to multiple networks that use SLAAC and embed the device's MAC address into the IPv6 address.
With this technique, the accuracy of models and outputs from processing the data set may be higher when used with IP networks that either assign publicly-accessible, non-RFC1918 IP addresses directly to the client, or when used with networks that use IPv6 addresses, considering that these configurations are less likely to use NAT or PAT, which obfuscates many clients behind a single set of IP addresses from the perspective of an external party. Likewise, the inferencing system ingesting data needs to be aware when processing events of the different network configurations of organizations since one organization may have 10,000 computers and route all their traffic through 10 IP addresses, while another has 100 computers routing traffic through 100 IP addresses. If using the distinct number of IP addresses as a proxy for volume, the variations between networks must be kept in mind. For example, in some cases, a single IP address is shared by several devices, so a certain number of transactions associated with that IP address may actually originate from more than one devices.
In some embodiments, the inferencing system determines the type of configuration that is used. To this end, the inferencing system may use another datasets to estimate or to obtain the number of machines/servers associated with a specific IP address for a given period of time. Alternatively, the system may analyze the semantics of the communication protocol and the expected traffic volume of the application, using which the number of servers associated with an IP address can be identified. For example, if an application communicates every 10 seconds with a certain amount of data transferred the number of server systems associated with that IP address can be determined by observing the frequency of communications and/or the amount of data transferred.
The IP addresses included in different data points that are associated with the same device are generally independent of each other. Thus, whether a device previously connected to an IPv4 network and subsequently connects to an IPv6 network or vice versa has no bearing on the integrity or quality of the data. If a device connects to two or more networks that support SLAAC for IPv6, however, the IP address itself may have a bearing on reducing noise in the observed data.
A network operator may be able to produce data that is suitable for this technique. The network operator may be required to produce data that conforms with the limitations on unnecessary events associated with RFC1918 IP addresses. As such, the data points having RFC1918 IP addresses solely, without publicly-routable IP addresses, may be excluded from the data set. Also, the vast majority of network operators generally only observe traffic to and from their respective networks, which may introduce bias, or a filter, on certain specific class of events without containing data points unrelated to the respective network operator's network. As such, typically the best provider(s) for flow information in this context include various transit providers that route traffic on behalf of clients, customers, peers, etc.
In the first set of transactions, that destination is ultimately the network systems operated, controlled, and/or used by “Business Partner, LLC.” The packet that was sent from the user device is routed across the Internet, typically through various other network operators and transit providers, as shown by the annotations “B1.” One of these transit providers logs the traffic transiting through their network, as shown at “C1,” where the logged traffic may include the transactions that the originating user device had sent.
Since routing on the Internet is not necessarily deterministic, it is possible that the route a packet ultimately took to “Business Partner, LLC” included the path indicated at “D1” and thus that packet bypassed the network operator who is saving flow information. It could also be the case that the packets sent to “Business Partner, LLC” go through “D1”, but packets returning to “Example, Inc.” may take the entire path as indicated by “B1” and thus “C1,” as well. Packets are exchanged between the user device and the server before the connection is ultimately terminated by one of the two parties.
In addition to communicating with the web system at “Business Partner, LLC,” the same user device at “Example, Inc.” and/or other user devices at “Example, Inc.” may communicate with the MTA by sending an email, as shown at “F1”, or send or receive a file using the on-premise solution at “Business Partner, LLC,” as indicated by “G1.” Each of these transactions, if routed entirely through “B1” may be logged as indicated at “C1.” The producer of the flow information may perform summary of the packets, or produce data points for each packet sent, as shown in
There exist many different protocols for which traffic can be observed, each serving distinct functions. While more and more higher-level application commonly use HTTP as the preferred method of transacting, there are still many services and applications that use other standard protocols that can provide visibility into the activity of certain devices on different networks. This differentiation and understanding of those services can be important given that the use of certain protocols (e.g. collaboration protocols like SMTP or XMPP, etc.) may imply greater evidence of a business relationship than others (e.g., the generic HTTP protocol. Some protocols exhibit certain patterns in flow information (e.g. common programmatic transactions may exhibit distinct or similar payloads or packets exchanged, from which the type of action that the device or user is performing can be inferred. Which particular protocol(s) is/are being used when the devices associated with one entity engage in transactions with another entity can be inferred from the port number(s) at the other entity that are accessed. Alternatively, or in addition, the protocol(s) can be identified by analyzing the flow information.
Another source of added-value for the inferencing system that analyzes this dataset includes data describing the context around individual IP addresses, where the context can be a factor in analyzing the existence of or nature of business relationships. For example, if a user interacts with an organization's primary website (e.g., “www.example.com”), hosted, for example, at IP address 192.0.2.203, with an arbitrary number of packets and volume exchanged, this data point alone may not specifically reveal a business relationship unless one understands the supported functions of the primary website. If those functions do not contain functions or information that a business partner may need to access, there may not be a strong evidence that this transaction represents a dependency or a relationship between the user's and the visited organizations.
The data set may include another IP address, e.g., 192.0.2.210, involved in one or more transactions with the user or user device associated with an entity. Contextual information about this other IP address, that may be unrelated to flow information, may reveal that this particular IP address exclusively hosts a web application tailored towards information exchange of that business's clients. This may be an important piece of information that provides strong evidence of a dependency or a relationship between the user's and the visited organizations. Other factors include frequency of visits, adjusted for the size of the originator's organization, so as to avoid a skew due to an organization having a large number of employees.
While
The fourth technique that a system identifying possible business relationships is based on the assumption and observation that users or devices, may have an implicit or explicit relationship or dependency with the organizations for which they are exchanging email as revealed by anti-spam and reputation technologies in services within the path of the email exchange. This technique uses data points containing information about the two organizations, such as described by either the IP addresses of the endpoint systems who are sending or receiving email, the MTAs who are processing email being sent or received, as well as the domains that are associated with the email accounts involved in the transaction.
Each data point in the data set required for this technique generally includes a timestamp which may be the point in time at which a transaction generating the data point is received by a downstream system, an identifier of one organization (domain component of an email address, IP address of the MTA, IP address of a DNS resolver, or IP address of the endpoint, etc.) and an identifier of another organization (domain component of an email address, IP address of the MTA, IP address of a DNS resolver, or IP address of the endpoint, etc.). The types of the two entity identifiers included in a data point can be but need not be the same. For example, the IP address of the MTA may be representative for one organization and a domain present may be representative for the other organization. Likewise, there may be multiple identifiers for one organization or the other, or both, within each one of the data points.
The combination of multiple types of attributes can be helpful when one or both of the organizations use an email service provider or third-party/external services firm to manage their email infrastructure. Under these conditions, the IP address of the respective MTAs would likely appear associated with the email service provider rather than the originating or destination organization. A reliable attribute would then be domain-level information as that is exclusive to each organization, unless they also happen to be using a generic email hosting provider, which is more common in small businesses. As such, the IP address of the endpoint, workstation, user device that sent or received the mail is likely a source of data leading to an accurate determination of the existent and/or nature of a relationship.
The data points used in this technique generally do not use RFC1918 IP addresses because such an IP address usually does not provide any notional value as a standalone attribute. An RFC1918 IP address may be provided alongside a publicly routed IP addresses where the RFC1918 IP address can provide value disambiguating between multiple clients behind a NAT. The IP address may be the form of an IPv4 address or an IPv6 address. An IPv6 address may allow for noise reduction if the device is connected to multiple networks that use SLAAC and embed the device's MAC address into the IPv6 address. However, unlike networks consisting of workstations, the IP addresses of MTAs are expected to remain very stable considering administrators often rely upon reputation of their previous activities to ensure reliable deliverability in the future.
The transactions that would generate the data points used by this technique typically occur in the realm of anti-spam and domain and IP reputation services. Operators of email infrastructure systems are incentivized to use a number of different opportunities and methods in order to reduce the amount of noisy email, or those that would be classified as spam, from being delivered to their end-users. As such they often employ various techniques during the entire email exchange lifecycle to obtain the probability that a given email that is in the process of being received is malicious or spam.
These transactions can occur at any point during the delivery process. Often, they are initiated by the MTAs owned and operated by the destination domain owner, or by the device or system of the end-user, or more specifically by the operator of the destination mailbox. The objective of these transactions is to gather as much external contextual information about the sender as possible, and as such, these technologies often employ investigating the domain associated with the sender's email address, investigating the entire mailbox associated with the sender's email address, investigating the IP address of the system connecting to the MTA(s) performing this investigation, investigating the content of the email address itself, etc.
During the course of programmatically discovering information about email addresses, domains, or IP addresses of the sender's system, the receiving MTA may provide certain information about itself to third-party reputation services or other systems. From this information, it is inferred in some embodiments that the entity associated with the MTA may have a business relationship with the entity associated with the sender of the email. As such, data points used in this technique may be generated by a reputation service and/or another system interacting with the reputation service, using the information received from the MTA or from another requestor requesting investigation of the email sender. For example, if a service is providing both domain and IP address reputation, a receiving MTA may be interested in querying the service for all inbound mail it is receiving. As such, it is providing this service with: (a) the domain being queried, which is the domain component of the sender's email address, i.e. the domain of the sender, (b) the IP address being queried, which is the IP address of the system sending the email, and (c) the IP address of the system receiving the email, or a proxy of that system that is performing the query. From this information, a business relationship can be inferred between the sender and the receiver entities. In some cases, such an inference may be drawn only when a certain threshold number of requests are observed.
These queries may be structured in many different ways and may include other contextual information from the exchange between the systems or from the email content or headers. The receiver system may use different channels to perform the quer(y/ies) depending on the implementation of the reputation service. For example, a query may be sent directly to the service via HTTP or some other protocol, or may be transmitted through another application. If the channel is DNS rather than direct HTTP transactions, the reputation information may be available beyond just the transactions. For example the reputation information may manifest within passive DNS feeds or other contexts where this information is public or shared (e.g. the reputation service may host a website detailing reputation of assets and the originating reports).
The receiver MTA may begin by constructing a query that investigates both the domain and IP address of the sender. This query may be formed as part of a single DNS query name and the query name may be suffixed with the one specified to be used by a particular anti-spam vendor. A suffix may be a function of the reputation service's API design. For example, if the reputation data can be queried via a DNS, the service may dictate that queries all take the form: <ip address>.<domain> with the service's domain name (e.g., “service-domain-name.com”) suffixed at the end. To illustrate, the domain “example.com” and the IP address “198.3.5.53” may be queried using the DNS query name “198.3.5.53.example.com.service-domain-name.com.” The receiver MTA forwards this query to the company's DNS resolver (“C1”) where it is queried throughout the DNS hierarchy before ultimately being sent to the anti-spam vendor's DNS authoritative name servers (“D1”) where the request is routed to the anti-spam vendor's internal service.
Once a reply is determined, it is sent back to the DNS resolver (“E1”), which is then replied back to the MTA operated by “Business Partner, LLC” (“F1”). Using input received from this anti-spam vendor, among many other possible interfaces and data sources, the MTA can pass that information into an internal model and filter to determine what action it may take on this individual email. Considering this is a legitimate business exchange, the MTA may pass the email message onto the recipient (“G1”) and into the recipient's mailbox.
In this example the MTA was the actor that initiated the interface with the anti-spam vendor, but any party in the path of the email exchange may interact with this vendor, either by a human initiated action or a programmatic one, as
The recipient may have its own anti-spam component on its local system. This is common within various anti-virus applications, that might also analyze the email origin and attachments within those emails. It is at this stage that the recipient user's system initiates the transaction with the DNS resolver to request reputation information (“D2”), possibly among other actions. Similar to the transactions in
The difference between these two scenarios can impact the characteristics of the queries and the volume observed for a given destination email domain. For example, when the receiving MTA performs anti-spam validation on all inbound mail to the organization for every mailbox, that generally affects more individual users than a single individual user performing it for his or her own mailbox. In both of these instances, the reputation and anti-spam vendors understand the context for which the data is being queried as well as the actors being queried and potentially where the query is originating. There can be other cases where other parties are aware of the frequency of reputation lookups on systems, including the actors being queried, as described below.
With reference to
The DNS resolver forwards this query to the anti-spam vendor (“D3”) and receives a response back based on the information included in the query (“E3”). Unlike the previous examples, the DNS resolver that “Example, Inc.” uses retains the queries and answers thereto as part of a DNS record retention or passive DNS platform. After it receives the answer back from the anti-spam vendor, it outputs this record to the passive DNS platform (“F3”). In some cases, the information stored in the passive DNS may be used to infer a business relationship, e.g., between Example, Inc. and Business Partner, LLC. The DNS resolver subsequently or simultaneously sends this record back to the receiving MTA (“G3”), which then passes this information onto the recipient's mailbox after determining that it is not malicious. The recipient user is then able to access the email, e.g., when the next time the user or a programmatic system checks the status of the mailbox (“H3”).
Technique 5: Inferencing Based on IP Address Geolocation AssociationThe fifth technique that a system identifying possible business relationships is based on the assumption and observation that users or devices, may have an implicit or explicit relationship or dependency with the organizations when the users or their devices are physically at another organization's office or other space used by that organization. It uses data points containing information about the two organizations, such as described by either the IP addresses of the endpoint system and or a device or user identifier, as well as the positioning information of the device.
Each data point in the data set required for this method generally includes a timestamp which may be the point in time at which a transaction generating the data point is received by a downstream system, an IP address of the device or an identifier of the user or device, as well as the physical location of the device. In the case when an IP address is absent from the dataset and a device identifier is present, multiple observations may be necessary to understand the context of the device and the organizations with which the device is associated.
With reference to
At other instances of time, similar observations may be made with respect to other user devices or users. For example, a device may be observed at a location associated with “Business Partner, LLC” and a user identifier “user1@email_provider.com” may be transmitted to an observer from the user device. At another time, a device may be observed at a location associated with “Example, Inc.” It may or may not be known whether the same device that was previously observed at the location of “Business Partner, LLC” is now observed at the location of “Example, Inc.” The same user identifier, “user1@email_provider.com” may be transmitted, however, to the observer when the user device is a the location of “Business Partner, LLC.”
From either sets of observations, i.e., those involving the device ID or the user ID, it may be inferred that a business relationship exists between “Example, Inc.” and “Business Partner, LLC.” The confidence in the inference can be improved by combining the two different types of data points.
There may exist other optional, value-added fields in one or more data points of the data set. An inferencing systems consuming this information can improve its accuracy by having both the device identifier and IP address available. Any of the known techniques to determine the current latitude and longitude data of the user device may be used. The contextual information on how the positioning information was derived, the precision, and/or positional error of that measurement can provide useful information about the quality of the data point. Information about the user device, or the type of device, can provide context on its expected behaviors and thus help eliminate or identify plausibly erroneous data points.
The data points used in this technique generally do not use RFC1918 IP addresses because such an IP address usually does not provide any notional value as a standalone attribute. An RFC1918 IP address may be provided alongside a publicly routed IP addresses where the RFC1918 IP address can provide value disambiguating between multiple clients behind a NAT. The IP address may be the form of an IPv4 address or an IPv6 address. An IPv6 address may allow for noise reduction if the device is connected to multiple networks that use SLAAC and embed the device's MAC address into the IPv6 address.
One may be able to derive additional device identifiers by observing user identifiers and vice versa. In general, this technique has no particular advantage when collecting data points from mobile devices versus from traditional workstations, or from laptops. There are, however, more instances of mobile devices in general, and it is more common for mobile devices to request and receive physical location information compared to other types of devices.
In some embodiments, only those data points that are mapped to a known physical address may be considered. Any information, not limited to only physical position or IP addresses, that is not associated with an entity may be ignored by the inferencing system analyzing those data points.
Analyzing Data PointsIn various embodiments, an inferencing system can consume data points produced by any of the above-described techniques. The system may obtain knowledge of the source from which each data point originated. This may allow the inferencing system to use the semantics of data gathering to calculate more accurate duration and currency information. For example, if one data set having data points collected over a month was known to show evidence of relationships, while another data set having data points collected over an hour was also known to show evidence of relationships, the inferencing system may not apply the same rules to the two data sets when determining whether an relationship remains current or active. This may be largely enforced by a downstream process internal within this system that is processing all known evidence of relationships and then determines whether they remain valid. A ruleset to be applied to a particular data set and/or to data received from a particular source may be defined statically, when configuring the system, or the rulesets can be defined dynamically, e.g., by applying machine learning models on the data on a continuous basis, where labeled data are available and indicate known relationships.
Each data point has a form of validity before and beyond its creation date. For example, a data point produced by Technique 1 may be produced immediately after a device connects to the network, or hours or days after it does. In either situation, the device and/or the user were present in the physical space associated with the entity providing, controlling, or using the network to which the user device connected for an unknown amount of time before the data point was produced, and likely remained there for an unknown amount of time after the data point was produced. Thus, in each of the techniques described above, each data point is a discrete measurement conveying some continuous behavior that lasted for an unknown duration when considered in isolation.
Regardless, the relationship between organizations can still remain active in the above example even when the individual and the device departs from the physical location of the network. The expected re-observation period generally varies by technique, but it may also vary based on how the data points were produced by the observing entity (e.g. a provider may cause a device to report beacons once a minute, another may be once a day or less, etc.). In some cases, it is desirable that the re-observation period is neither too long, otherwise false positives can be introduced, nor too short, otherwise false negatives could appear. The re-observation period may also be a function of the entity size or other entity attributes. Thus, each data point has an extension of validity both before and after it was created. The lifetime of these data points can vary and it is up to the interpreter to understand what those may be.
Using relationships that are known to exist between two entities, the techniques described herein can allow an interpreter to understand how the inferencing may vary by data set, analysis techniques, and/or by provider of data sets. Size of organization, type of relationship, etc., among other differentiating factors may also be considered. This analysis may be performed on an on-going basis, e.g., to ensure that the system outputs results that at least include the known relationships. Likewise, false positives can be reduced by identifying behaviors in the techniques and datasets that lead to relationships being identified that are known not to exist.
With reference to
The first stage involves processing the actual data points to analyze their meaning and optionally map the data points to a common form (#1). For example, each data set may not be delivered to the inferencing system in the same form/format or schema. One data set may be encoded as a CSV, another may be JSON objects, another may be a byte stream. The inferencing system may ingest and interpret these different forms of data sets using suitable parsers and/or interpreters.
The data may be de-duplicated to reduce the number of data points consumed in subsequent processing. A summary of the deduplication information may be embedded into a single event. For example, in the case of a data points originating from a data set produced according to technique 1, there may have been many data points produced by a specific device in a given transaction. These data points may be reduced down to a single data point with the count of data points and a range of timestamps for which that data point was observed. The dataset may then be written into an intermediate form into an arbitrary destination (#2).
Care is taken when performing deduplication, as the time series for which deduplication is performed may be too large, and loss of valuable information may occur. For example, in the case of a dataset produced according to Technique 1, if a device was on a network a specific point in time, transitioned to another network, and then returned to the first network, then the first and third event may not be deduplicated unless the system is able to retain the two distinct states within a combined event resulting from the deduplication.
The next stage (#3) is that once the system understands the data point, it may then enrich the data points with as much other information that is available or accessible to the system. If the inferencing system decides to retain a particular data point as evidence of a relationship, or decides to transform that data point into information that describes an inferred relationship, the system may optionally store this information with the data point. This can include, but is not limited to the following.
Enriching IP addresses to include network characteristic information (#4), such as but not limited to, classification of the IP address, whether it be residential, cellular, or other form that is relevant; known and associated wireless access points, which may be indicative of whether that specific IP address is known to be hosting a guest network, either shared or exclusive; known activity on that IP address; etc.
Enriching domains (#5) to include the company that is responsible for that asset at the time the data point was created.
Enriching IP addresses with current or historical devices associated with that network (#6), whether they are endpoint-machines like workstations or mobile phones or infrastructure providing services, etc.
Enriching IP addresses to include known physical location information associated with the IP address (#7) if the data point lacked that information, whether it is a standard geolocation conversion to a country, state, city, neighborhood or if it is to associated precision location information. This also includes translating precision location information into a set of plausible street addresses. This can be done using many services not described herein. These street addresses may then be joined against a repository of known street addresses belonging to various businesses.
Enriching IP addresses to include the company that is responsible for that asset at the time of the data point was created (#8), as specified by the timestamp of the data point. This is necessary to ultimately create a relationship between one company and another organization. This is also important to understand as a company uses many different IP addresses, so in some respects it can be thought as a deduplication effort. For example, if an organization owns and operates two different IP addresses, a data point associated with one IP address and a data point associated with the second data point can be treated as being associated to the same entity. This is particularly important when considering lifetime and durations of relationships, since it should not be expected to re-observe a relationship to the same individual type of evidence that was used to originally create the relationship.
Data points may also be considered time series of data points in the same data set or set of known relations (#9), or data points derived from different data sets outside of those processed by the system (#10). Each example provided above showed data points as represented as a time series. When looking at a single data point, it will generally provide value knowing what has preceded it. As such, part of the enrichment may include looking at other data points associated with one or more fields of the data points that is being assessed. For example, it may be beneficial to have all other data points associated with the IP address on hand when analyzing the current data point that is associated with that IP address and that is being processed.
The third stage is to determine if there is any value in the data point that is being received (“Analytics Engine”). A data point may be erroneous outright, for example due to requirements in the precision of the information received such as the data points generated according to Technique 5 that processes geographic data points, one would not want to consume (i.e., process or analyze) data points that have a large horizontal positioning error. Likewise, as discussed in each section about each method, there may be certain requirements that are desired, or a certain amount of confirmation that is necessary before the system should conclude that a relationship is current given a certain threshold of false positives.
A data point may be deemed to be erroneous but can also still provide value. For example, if a relationship is known to exist between two companies due to a series of previously processed data points, a data point that still proves this relationship is current, but has been deemed to have a higher margin of error, may still be useful to affirm that the relationship is current. As such, there may exist different levels of tolerance for errors, inaccuracies, attributes of false positives, or other function, that would deem a data point to be more reliable based on the task at hand, such as using the data point to: (i) reveal that a new relationship is present, (ii) re-enforce an existing relationship, or (iii) terminate a known relationship.
The fourth stage includes writing that information back into a respective datastore (#11), which can be a messaging queue or a formalized database or a traditional file store, depending on the requirements of any application interested in consuming information produced by such a system.
This system may simply output evidence (e.g. relationships between two IP addresses), and leave it up to another process to read all events, a portion of events (such as those still considered to be relevant for today versus when they were created), and to conclude that the entities associated with the two IP addresses, where the entities may be identified, e.g., by mapping the IP addresses to entities, have a relationship.
These steps describe how an inferencing system can process existing data. A part of the system may also process historical evidence of relationships to determine whether any previously inferred relationships remain active. This function may be left up to the consumers of the data, e.g., based on the tolerance for false positives. Entities generally vary in number of employees, number of active devices, and many other characteristics that can cause different entities to be represented differently in each of the data sets derived from the different techniques described herein. As such, a downstream consumer of relationships that has a higher tolerance for false positives may want to consume relationships that are known to exhibit features that indicate a higher likelihood of being a false positive (e.g., the relationship was not re-observed in any dataset, or the threshold for which a relationship could be conclusive has not been reached).
If an entity asset (such as, but not limited to, IP address, domain name, etc.) is not known at the time when processing a data point, but in the future it is understood that the ownership of that asset covered the period of time that the data point was produced, then it would be valuable to infer any relationships involving that entity in the future. It may be the case that one of the two owners are not known, or perhaps both of the owners are not known. As such, one way to embed relationship information is by its evidence such that a downstream consumer of these data points can process the current and historical data points to know which of the entities were/are associated with each other.
The inference of business relationships may be used directly, e.g. to assess the strength of cybersecurity of an entity in the context of its partners. Other indirect uses of the inferences include determining the quality of data sets indicative of such relationships. For example, if through the aforementioned techniques a relationship was identified between a company and a local airport, e.g., based on employees of the company traveling through that airport, this is an inference that can be used directly. However, if this relationship ceases to exist or is significantly diminished over a period of time, while the volume of the company employees at another airport significantly increases, and no other public information is known about this company, then it may be reasonable to assume that the observed company has moved its office to a different location. In this case, other data sets may not have been updated to demonstrate the change in the physical office of the observed company.
Some uses of these techniques permit a higher tolerance for false positives. For example, when recommending a vendor to be added to a portfolio of companies for the purposes of monitoring the security posture of the companies in the portfolio, such a recommendation engine has a notable tolerance for false positives, although it cannot be too high or else the user interested in the security profile may not find substantial value from such an inference. On the other hand, if an organization's security posture is understood to be a function of its relationships, then the tolerance for false positives may be minimal to non-existent, as one would not want an erroneous false positive causing an inaccurate representation of that organization's security posture.
In some examples, some or all of the processing described above can be carried out on a personal computing device, on one or more centralized computing devices, or via cloud-based processing by one or more servers. In some examples, some types of processing occur on one device and other types of processing occur on another device. In some examples, some or all of the data described above can be stored on a personal computing device, in data storage hosted on one or more centralized computing devices, or via cloud-based storage. In some examples, some data are stored in one location and other data are stored in another location. In some examples, quantum computing can be used. In some examples, functional programming languages can be used. In some examples, electrical memory, such as flash-based memory, can be used.
A computing system used to implement various embodiments may include general-purpose computers, vector-based processors, graphics processing units (GPUs), network appliances, mobile devices, or other electronic systems capable of receiving network data and performing computations. A computing system in general includes one or more processors, one or more memory modules, one or more storage devices, and one or more input/output devices that may be interconnected, for example, using a system bus. The processors are capable of processing instructions stored in a memory module and/or a storage device for execution thereof. The processor can be a single-threaded or a multi-threaded processor. The memory modules may include volatile and/or non-volatile memory units.
The storage device(s) are capable of providing mass storage for the computing system, and may include a non-transitory computer-readable medium, a hard disk device, an optical disk device, a solid-date drive, a flash drive, or some other large capacity storage devices. For example, the storage device may store long-term data (e.g., one or more data sets or databases, file system data, etc.). The storage device may be implemented in a distributed way over a network, such as a server farm or a set of widely distributed servers, or may be implemented in a single computing device.
The input/output device(s) facilitate input/output operations for the computing system and may include one or more of a network interface devices, e.g., an Ethernet card, a serial communication device, e.g., an RS-232 port, and/or a wireless interface device, e.g., an 802.11 card, a 3G wireless modem, or a 4G wireless modem. In some implementations, the input/output device may include driver devices configured to receive input data and send output data to other input/output devices, e.g., keyboard, printer and display devices. In some examples, mobile computing devices, mobile communication devices, and other devices may be used as computing devices.
In some implementations, at least a portion of the approaches described above may be realized by instructions that upon execution cause one or more processing devices to carry out the processes and functions described above. Such instructions may include, for example, interpreted instructions such as script instructions, or executable code, or other instructions stored in a non-transitory computer readable medium.
Various embodiments and functional operations and processes described herein may be implemented in other types of digital electronic circuitry, in tangibly-embodied computer software or firmware, in computer hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of one or more of them. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible nonvolatile program carrier for execution by, or to control the operation of, data processing apparatus. Alternatively or in addition, the program instructions can be encoded on an artificially generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus. The computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.
The term “system” may encompass all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, or multiple processors or computers. A processing system may include special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit). A processing system may include, in addition to hardware, code that creates an execution environment for the computer program in question, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of them.
A computer program (which may also be referred to or described as a program, software, a software application, a module, a software module, a script, or code) can be written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages, and it can be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program may, but need not, correspond to a file in a file system. A program can be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.
The processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).
Computers suitable for the execution of a computer program can include, by way of example, general or special purpose microprocessors or both, or any other kind of central processing unit. Generally, a central processing unit will receive instructions and data from a read-only memory or a random access memory or both. A computer generally includes a central processing unit for performing or executing instructions and one or more memory devices for storing instructions and data. Generally, a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto optical disks, or optical disks. However, a computer need not have such devices. Moreover, a computer can be embedded in another device, e.g., a mobile telephone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a Global Positioning System (GPS) receiver, or a portable storage device (e.g., a universal serial bus (USB) flash drive), to name just a few.
Computer readable media suitable for storing computer program instructions and data include all forms of nonvolatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto optical disks; and CD-ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.
To provide for interaction with a user, embodiments of the subject matter described in this specification can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user and a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input. In addition, a computer can interact with a user by sending documents to and receiving documents from a device that is used by the user; for example, by sending web pages to a web browser on a user's user device in response to requests received from the web browser.
Embodiments of the subject matter described in this specification can be implemented in a computing system that includes a back end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front end component, e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the subject matter described in this specification, or any combination of one or more such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include a local area network (“LAN”) and a wide area network (“WAN”), e.g., the Internet.
The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
While this specification contains many specific implementation details, these should not be construed as limitations on the scope of what may be claimed, but rather as descriptions of features that may be specific to particular embodiments. Certain features that are described in this specification in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable sub-combination. Moreover, although features may be described above as acting in certain combinations and even initially claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a sub-combination or variation of a sub-combination.
Similarly, while operations are depicted in the drawings in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.
Particular embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. For example, the actions recited in the claims can be performed in a different order and still achieve desirable results. As one example, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In certain implementations, multitasking and parallel processing may be advantageous. Other steps or stages may be provided, or steps or stages may be eliminated, from the described processes. Accordingly, other implementations are within the scope of the following claims.
The phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. The term “approximately”, the phrase “approximately equal to”, and other similar phrases, as used in the specification and the claims (e.g., “X has a value of approximately Y” or “X is approximately equal to Y”), should be understood to mean that one value (X) is within a predetermined range of another value (Y). The predetermined range may be plus or minus 20%, 10%, 5%, 3%, 1%, 0.1%, or less than 0.1%, unless otherwise indicated.
The indefinite articles “a” and “an,” as used in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean “at least one.” The phrase “and/or,” as used in the specification and in the claims, should be understood to mean “either or both” of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with “and/or” should be construed in the same fashion, i.e., “one or more” of the elements so conjoined. Other elements may optionally be present other than the elements specifically identified by the “and/or” clause, whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, a reference to “A and/or B”, when used in conjunction with open-ended language such as “comprising” can refer, in one embodiment, to A only (optionally including elements other than B); in another embodiment, to B only (optionally including elements other than A); in yet another embodiment, to both A and B (optionally including other elements); etc.
As used in the specification and in the claims, “or” should be understood to have the same meaning as “and/or” as defined above. For example, when separating items in a list, “or” or “and/or” shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly indicated to the contrary, such as “only one of or “exactly one of,” or, when used in the claims, “consisting of,” will refer to the inclusion of exactly one element of a number or list of elements. In general, the term “or” as used shall only be interpreted as indicating exclusive alternatives (i.e. “one or the other but not both”) when preceded by terms of exclusivity, such as “either,” “one of,” “only one of,” or “exactly one of” “Consisting essentially of,” when used in the claims, shall have its ordinary meaning as used in the field of patent law.
As used in the specification and in the claims, the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, “at least one of A and B” (or, equivalently, “at least one of A or B,” or, equivalently “at least one of A and/or B”) can refer, in one embodiment, to at least one, optionally including more than one, A, with no B present (and optionally including elements other than B); in another embodiment, to at least one, optionally including more than one, B, with no A present (and optionally including elements other than A); in yet another embodiment, to at least one, optionally including more than one, A, and at least one, optionally including more than one, B (and optionally including other elements); etc.
The use of “including,” “comprising,” “having,” “containing,” “involving,” and variations thereof, is meant to encompass the items listed thereafter and additional items. Use of ordinal terms such as “first,” “second,” “third,” etc., in the claims to modify a claim element does not by itself connote any priority, precedence, or order of one claim element over another or the temporal order in which acts of a method are performed. Ordinal terms are used merely as labels to distinguish one claim element having a certain name from another element having a same name (but for use of the ordinal term), to distinguish the claim elements.
Claims
1-21. (canceled)
22. A method for inferring a relationship between two entities, the method comprising the steps of:
- receiving at a server, from a network device, composite flow information corresponding to a plurality of flows, wherein each individual flow information comprises a source network identifier and a destination network identifier;
- determining that for a subset of the plurality of flows: (i) the source network identifier in each flow in the subset belongs to a first set of network identifiers, each of which being associated with a first entity, and (ii) the destination network identifier in each flow in the subset belongs to a second set of network identifiers, each of which being associated with a second entity; and
- determining that a relationship exists between the first entity and the second entity based on, one or more of: (i) a total number of flows in the subset, (ii) a frequency of the flows in the subset, (iii) a total size of the flows in the subset, (iv) a port associated with the flows in the subset.
23. The method of claim 1, wherein determining the existence of the relationship comprises determining that:
- the total number of the flows in the subset is at least equal to a specified flow-count threshold; or
- the frequency of the flows in the subset is at least equal to a specified flow-frequency threshold; or
- the total size of the flow in the subset is at least equal to a specified flow-size threshold.
24. The method of claim 2, wherein the flow-count threshold, the flow-frequency threshold, or the flow-size threshold is based on, at least in part, a size of the first entity or a size of the second entity.
25. The method of claim 1, wherein determining the existence of the relationship comprises identifying a type of a port associated with the subset of flows.
26. The method of claim 4, wherein the port type is a file transfer protocol (FTP) port, or a simple mail transfer protocol (SMTP) port.
27. The method of claim 1, wherein the determination of existence of the relationship is based on, at least in part, an additional determination that one or more of the network identifiers in the second set are designated for an entity having a relationship with the second entity.
28. The method of claim 1, wherein the network device is associated with an Internet service provider (ISP) or an Internet exchange point (IXP), the ISP or the IXP being different from the first entity and the second entity.
29. The method of claim 1, wherein:
- the network device comprises a domain name system (DNS) resolver; and
- a first individual flow information comprises a first source network identifier, a first destination network identifier, and a response from a reputation service corresponding to the first source network identifier.
30. A system for inferring a relationship between two entities, comprising:
- a processor;
- a network port in communication with the processor and adapted to receive composite flow information corresponding to a plurality of flows; and
- a memory coupled to the processor and comprising instructions, which when executed by the processor, program the processor to: receive from a network device, the composite flow information corresponding to a plurality of flows, wherein each individual flow information comprises a source network identifier and a destination network identifier; determine that for a subset of the plurality of flows: (i) the source network identifier in each flow in the subset belongs to a first set of network identifiers, each of which being associated with a first entity, and (ii) the destination network identifier in each flow in the subset belongs to a second set of network identifiers, each of which being associated with a second entity; and determine that a relationship exists between the first entity and the second entity based on, one or more of: (i) a total number of flows in the subset, (ii) a frequency of the flows in the subset, (iii) a total size of the flows in the subset, (iv) a port associated with the flows in the subset.
31. The system of claim 30, wherein to determine the existence of the relationship, the instructions program the processor to determine that:
- the total number of the flows in the subset is at least equal to a specified flow-count threshold; or
- the frequency of the flows in the subset is at least equal to a specified flow-frequency threshold; or
- the total size of the flow in the subset is at least equal to a specified flow-size threshold.
32. The system of claim 31, wherein the flow-count threshold, the flow-frequency threshold, or the flow-size threshold is based on, at least in part, a size of the first entity or a size of the second entity.
33. The system of claim 30, wherein to determine the existence of the relationship, the instructions program the processor to:
- identify a type of a port associated with the subset of flows.
34. The system of claim 33, wherein the port type is a file transfer protocol (FTP) port, or a simple mail transfer protocol (SMTP) port.
35. The system of claim 30, wherein to determine the existence of the relationship, the instructions program the processor further to:
- determine that one or more of the network identifiers in the second set are designated for an entity having a relationship with the second entity.
36. The system of claim 30, wherein the network device is associated with an Internet service provider (ISP) or an Internet exchange point (IXP), the ISP or the IXP being different from the first entity and the second entity.
37. The system of claim 30, wherein:
- the network device comprises a domain name system (DNS) resolver; and
- a first individual flow information comprises a first source network identifier, a first destination network identifier, and a response from a reputation service corresponding to the first source network identifier.
Type: Application
Filed: Aug 23, 2019
Publication Date: Feb 25, 2021
Patent Grant number: 11956265
Inventor: Daniel Dahlberg (Somerville, MA)
Application Number: 16/549,764
